neural_network

neural_network, an Octave code which illustrates the use of a neural network for deep learning, using back propagation and stochastic gradient descent, by Catherine Higham and Desmond Higham.

Licensing:

The information on this web page is distributed under the MIT license.

Languages:

neural_network is available in a MATLAB version and an Octave version and a Python version.

Related Data and Programs:

neural_network_test

Author:

Catherine Higham, Desmond Higham. Modifications by John Burkardt.

Reference:

1. Catherine Higham, Desmond Higham,
   Deep Learning: An introduction for applied mathematicians,
   SIAM Review,
   Volume 61, Number 4, pages 860-891,
   November 2019.

Source Code:

• activate.m, evaluates the sigmoid function y = sigmoid(W*x + b).
• cost_function.m, evaluates the cost function.
• cost_plot.m, plots the cost function over the iterations.
• intspace.m, returns roughly equally spaced integers in an interval.
• neural_network.m, uses back propagation to train a network.